On 2/5/22 04:35, Peter Maydell wrote:
On Fri, 4 Feb 2022 at 07:53, Richard Henderson
<richard.hender...@linaro.org> wrote:
This will allow us to control exactly what scratch register is
used for loading the constant. Also, fix a theoretical problem
in recursing through tcg_out_movi, which may provide a different
value for in_prologue.
Signed-off-by: Richard Henderson <richard.hender...@linaro.org>
---
tcg/sparc/tcg-target.c.inc | 20 ++++++++++++--------
1 file changed, 12 insertions(+), 8 deletions(-)
static void tcg_out_movi(TCGContext *s, TCGType type,
TCGReg ret, tcg_target_long arg)
{
- tcg_out_movi_int(s, type, ret, arg, false);
+ tcg_debug_assert(ret != TCG_REG_T2);
+ tcg_out_movi_int(s, type, ret, arg, false, TCG_REG_T2);
Here we assert that 'ret' isn't TCG_REG_T2, but in
tcg_out_addsub2_i64() we do:
tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_T2, bh);
and
tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_T2, bh + (is_sub ? -1 : 1));
Otherwise looks OK.
Oops. Good catch. I shouldn't have moved the assert.
r~
